DRBD too slow to start with software RAID

Hello Yall,

I'm starting to play with DRBD to build a small (2-node)
failsafe cluster here.

The problem is, although DRBD starts really quickly with my
SATA
drive, with the MD (software RAID) device it's REALLY slow.

It's like:

Dec  7 18:41:12 mail0 drbd0: resync bitmap: bits=87603792
words=2737620
Dec  7 18:41:12 mail0 drbd0: size = 334 GB (350415168 KB)
Dec  7 18:44:49 mail0 drbd0: 0 KB marked out-of-sync by on
disk bit-map.
Dec  7 18:44:50 mail0 drbd1: resync bitmap: bits=50975584
words=1592988
Dec  7 18:44:50 mail0 drbd1: size = 194 GB (203902336 KB)
Dec  7 18:44:51 mail0 drbd1: 0 KB marked out-of-sync by on
disk bit-map.

DRBD0 is built over /dev/md4 that is built over two SATA
disks.

While this load process is running, I tried a dstat -df to
see what
disk activity I'm having and I found out that DRBD was
reading at
100Kbps.

On the other server it happens exactly otherwise. That's
because the
DRBD0 is on the SATA and DRBD1 is on the MD4:

Dec  7 18:41:12 www0 drbd0: resync bitmap: bits=87603792
words=2737620
Dec  7 18:41:12 www0 drbd0: size = 334 GB (350415168 KB)
Dec  7 18:41:14 www0 drbd0: 0 KB marked out-of-sync by on
disk bit-map.
Dec  7 18:41:14 www0 drbd1: resync bitmap: bits=50975584
words=1592988
Dec  7 18:41:14 www0 drbd1: size = 194 GB (203902336 KB)
Dec  7 18:43:19 www0 drbd1: 0 KB marked out-of-sync by on
disk bit-map.

And that proves that it is a problem with the
DRBD-SoftwareRAID couple.

Versions:
Kernel: 2.6.18-gentoo-r3
DRBD: 0.7.22 (although it's not on Portage, it's the only
one that
works with kernel 2.6.18).

Thank you all.

Best regards,
Daniel Colchete
-- 
gentoo-clustergentoo.org mailing list

Hello Yall,

I found a way to by-pass the problem, so I'm sending it to
the list so
that it's archived.

It's a problem between the software raid devices and DRBD's
metadata.
My metadata was internal. So, the solution is to put the
metadata out
of the /dev/md? device. Do it and it'll start in one second.
I didn't
try to put the metadata on a separate MD device.

The problem is that with the metadata out of the RAID1, then
you have
a problem if the metadata harddrive fails. That's why I'm
still
persuing this.

Best regards,
Daniel Colchete

On 12/7/06, Daniel van Ham Colchete <daniel.colchetegmail.com> wrote:
> Hello Yall,
>
> I'm starting to play with DRBD to build a small
(2-node) failsafe cluster here.
>
> The problem is, although DRBD starts really quickly
with my SATA
> drive, with the MD (software RAID) device it's REALLY
slow.
>
> It's like:
>
> Dec  7 18:41:12 mail0 drbd0: resync bitmap:
bits=87603792 words=2737620
> Dec  7 18:41:12 mail0 drbd0: size = 334 GB (350415168
KB)
> Dec  7 18:44:49 mail0 drbd0: 0 KB marked out-of-sync by
on disk bit-map.
> Dec  7 18:44:50 mail0 drbd1: resync bitmap:
bits=50975584 words=1592988
> Dec  7 18:44:50 mail0 drbd1: size = 194 GB (203902336
KB)
> Dec  7 18:44:51 mail0 drbd1: 0 KB marked out-of-sync by
on disk bit-map.
>
> DRBD0 is built over /dev/md4 that is built over two
SATA disks.
>
> While this load process is running, I tried a dstat -df
to see what
> disk activity I'm having and I found out that DRBD was
reading at
> 100Kbps.
>
> On the other server it happens exactly otherwise.
That's because the
> DRBD0 is on the SATA and DRBD1 is on the MD4:
>
> Dec  7 18:41:12 www0 drbd0: resync bitmap:
bits=87603792 words=2737620
> Dec  7 18:41:12 www0 drbd0: size = 334 GB (350415168
KB)
> Dec  7 18:41:14 www0 drbd0: 0 KB marked out-of-sync by
on disk bit-map.
> Dec  7 18:41:14 www0 drbd1: resync bitmap:
bits=50975584 words=1592988
> Dec  7 18:41:14 www0 drbd1: size = 194 GB (203902336
KB)
> Dec  7 18:43:19 www0 drbd1: 0 KB marked out-of-sync by
on disk bit-map.
>
> And that proves that it is a problem with the
DRBD-SoftwareRAID couple.
>
> Versions:
> Kernel: 2.6.18-gentoo-r3
> DRBD: 0.7.22 (although it's not on Portage, it's the
only one that
> works with kernel 2.6.18).
>
> Thank you all.
>
> Best regards,
> Daniel Colchete
>
-- 
gentoo-clustergentoo.org mailing list